Autonomous Navigation of Drones


In the final project of the robotics course at the National Taiwan University, we addressed autonomous path planning and tracking in a 3D space using an aircraft. Our testing environment included two rings, that were recognized by a quadraotor's front camera, which than flew through them. We applied our results in an AR.Drone V1 application. This drone became popular in recent years for educational purposes in robotics. We used shape detection, dynamic navigation planning, and object depth calculation to accomplish the task. We introduced an approach, which expresses the depth of a landmark by a linear system, only using multiple pictures of the same scene. Approximating this linear system by the least square method resulted in accurate depth calculation. Besides challenges mentioned above, we had to face unstable control of slow and percise movements and inaccurate sensing.

This work was a team project realized together with my class mates Cristan Gonzalez, Holger Rasmussen und Barbara Sepic.